基于无人机图像分析的风机叶片表面损伤检测

Surface Damage Detection of Wind Turbine Blades Based on UAV Image Analysis

风机叶片是风电巡检的重要对象。针对传统人工检测效率低下且准确率低以及基于深度学习方法所面临的数据集制作困难等问题,提出了一种基于无人机图像分析的风力发电装置叶片表面损伤检测方法。首先,采用非局部均值算法对图像进行去噪处理,并采用局部二值模式提取样本的纹理特征。其次,提高支持向量机的分类能力,使用灰狼算法对支持向量机的惩罚因子和核函数参数进行寻优。最后,采用经过灰狼算法优化的支持向量机对样本的局部二值模式特征进行训练,得到风机叶片表面损伤检测模型。该方法识别正确率达96.2%。理论分析及实验结果表明,该方法检测成本低廉,可靠性高,推广性强。

The wind blades were the important targets for patrol inspection. In view of the low efficiency and accuracy of the traditional artificial inspection and the difficulty in making data sets based on the deep learning method, a damage inspection method based on the image analysis of the unmanned aerial vehicle was proposed. First of all, the non local average method was used to remove the noise from the image, and the local two value mode was used to extract the texture features of the samples. Secondly, it could improve the ability of the classification of the machines. And it could also find the best punishment factors and the core function of the machines with the grey wolf formula. In the end, the features of the local two values mode of the sample were trained with the support of the grey wolf formula, and the damage detection model on the surface of the leaf was obtained. The accuracy of this method was 96.2%. The theoretical analysis and experiment results showed that this method was cheap, reliable and easy to be promoted.